Reference based motion correction vs. Relion polishing

Hi all,
The new RBMC is fantastic. It works really well in the case the protein is of reasonable size but I found the result is a bit different vs. Relion polishing in a very big assembly.

I worked on the doublet microtubule which is very large. When I compared RBMC vs. Relion Polishing maps, the quality of RMBC map is equivalent as Relion Polished map at the region with the best resolution. However, the quality is of RMBC map in the worse region look a lot worse, a lot of smearing. And that doesn’t change when I run local refinement of the worse region. With relion polished particles, the appearance looks a lot better (using cryosparc to refine).

It is clearer in the snapshot image of the map. With a lot more smearing on the outside vs. the middle in RMBC vs. Relion polished map.

Relion polished particles, refined in cryosparc 3.5A

RMBC, refined in cryosparc 3.5A

Is there any big difference in the implementation of RMBC vs. Relion? If I used the hyperparamter found in Relion Polishing and used for RMBC, would I get similar result?

Hi @builab,

Interesting! Thanks for discovering this. There are some differences between Relion’s Polishing and CryoSPARC’s RBMC, and hyperparameters are not transferrable between them. Could you please share:

  • the box size, pixel size, and raw data format (tiff, eer, etc)
  • the following plots from the FCC job:
    • FCC plot from dose weight optimization
    • An example trajectories plot from the final motion correction stage
    • The trajectory activity plot from the last hyperparameter search iteration

– Harris